Ep 29: This episode: Variable Pricing & Conversational AI

Upside/Downside is a podcast for finance and business geeks about the hard work of value creation with a dose of humor thrown in for good measure.  I'm your host, Matt Cooley. 

Panelists Dana Price and Sami Akbay join me to discuss variable pricing models, and why we should expect wider adoption going forward.  Is P x Q just a distant dream?

Then we turn our attention to conversational artificial intelligence.  As the technology moves into a generative phase, will there be anything left for humans to do at the office?

Wonky, funny, thought-provoking discussion waiting for you on this episode of Upside/Downside.

Thank you for listening and please visit Upside/Downside podcast and enter your email for my FREE list: "10 places to look for higher profits and cash flow right now!".

I wish you the best on your value creation journey and if you find yourself stuck or in need of advice, please reach out via the contact form at Upside/Downside podcast today!

Matt

Transcript

SPEAKER_02: 0:01

Welcome back, everyone. This is Matt Cooley, host of Upside Downside, where our panel explores the value creation angle of current news stories and how the actions we take affect profits and cash flow. Occasionally, we also delve into our own personal psychoses, but we do try to keep that to a minimum. By day, I'm the head of finance for Ericsson's global network platform API business and a self-professed nerd for value creation and how it impacts companies and everyday people. Joining me are distinguished panelists panelists, Dana Price and Sami Akbe. Dana is CFO at Area 9 Lyceum Group and never lets her burn rate get out of control. She also breaks technology like last week when her neighbor saw Dana back the car over an old Commodore 64. Welcome, Dana.

SPEAKER_01: 0:48

Thanks. It's great to be here, Matt.

SPEAKER_02: 0:53

Sami Akbe is a technology executive and founder and is one of those people who knows how to shove data into the black box and make something useful come out the other side. He painted the words Mr. Data on the back of his shoes so people know that he's serious. Welcome, Sami.

SPEAKER_00: 1:09

Well, thank you for having me here, Matt.

SPEAKER_02: 1:11

Do you have those shoes on?

SPEAKER_00: 1:12

No, I'm wearing green shoes

SPEAKER_02: 1:15

today. Okay. Today we're going full wonky and discussing variable pricing models and conversational AI technology. I'm sitting here thinking, does this mean I have to haggle with chatbots for Sixers tickets? But we'll get into that later. So let's take variable pricing For quite some time now, companies have been offering variable or dynamic prices for airline seats, hotel rooms, ride shares, sporting events, concerts, and so on to consumers. And in the world of wholesale and commodities, it's been going on forever, right? What's evolving is the use of data analytics to create algorithms that optimize for a lot more than just price times quantity. Factors like your personal demographics, your physical location, the operating system on your smartphone, how many times you searched a keyword string in the last week even, all that stuff's being used to tune and present sort of hyper-optimized prices to buyers. What does this do to value creation and are there longer-term risks we should be thinking about? I mean, there's a lot here. Dana, you're a serious sports fan and a CFO. What are your observations about the upsides and downsides of variable pricing?

SPEAKER_01: 2:29

Sure. So from an upside perspective, I think there's a couple of things. There's potentially more predictable revenue streams, clearly the ability to perhaps make more money, and I'm thinking about it from the company's perspective, obviously, and also better staffing, better way to serve the support services for that model. On the flip side of that, from a consumer's perspective, I know that I am absolutely going to pay more for this muffler, George, at a certain point in time if I go to whether it be an event or whatever the related item is in the off hours. And there's two specific things where I can actually show you there are folks at the opposite ends. So I was at a conference once where the data guy from the Boston Red Sox was talking about how they implemented their pricing model. And this is several years ago. And they basically just turned all the data and looked at specific day of the week, time of the season, time of the game, and then obviously location of the seat in the ballpark to drive the pricing, i.e. when the Yankees play the Red Sox on a Saturday night, it's going to cost you a fortune. But on the flip side of that, they were also able to determine their staffing levels a little bit better, understand the predictability of what the food and beverage might run. They their food and beverage revenue so that they could staff the, you know, the counters and stuff on certain games. And he also did make a note, and I presume he was referring to when the Yankees play the Red Sox, the need for security and or safety staff. Apparently he has data on that where the, and this is probably a direct correlation between an increase in the food and beverage relates to an increase in the need for security

SPEAKER_02: 4:30

and safety. Beer consumption directly correlates to the need for security, something like that?

SPEAKER_01: 4:36

Yes, and apparently there's very specific places in the park where there's more security risk. So I think from a security perspective, that's actually pretty interesting. On the complete opposite end of the spectrum is Mercedes-Benz Stadium in Atlanta, and this was pre-COVID. I can't speak to it today, but I presume it's similar. Their data person basically ran all the data from all the events and determined that if they cut their prices by 50%, they would actually get more money per customer. And what that means is I can speak to a football event. So if you go to see the Atlanta Falcons play, there were five of us for lunch, and it cost only$32. The tickets, I believe, were something like$35. Now, the Atlanta Falcons weren't doing very well. Well, that year, but that doesn't seem to have an impact. The fact of the matter is that I did go back and I did get the extra drink or I did go buy a hat or I did buy more because I'm like, oh, it's only$20. So I'll just go buy something else. So you have complete opposite ends of the spectrum from a variable pricing perspective that is going to drive consumers in various ways, depending upon how they feel, whether they're being taken advantage of or, oh, this is a bargain.

SPEAKER_02: 5:59

Wow, that's fascinating. Fascinating. And to back it up with data like that, I wonder if any other teams have taken similar

SPEAKER_01: 6:07

approaches. I haven't been to Philly. I mean, I live two hours from there, but I haven't been to Philly yet. And I'm certainly not going to go next year because the Eagles were in the Super Bowl. So that's going to cost me more. Whereas if I go up to Gillette Stadium, where the Patriots play, it's going to cost me so much less than it would have a few years ago.

SPEAKER_02: 6:52

I love how you've broken this down to your personal entertainment matrix here. I'm sorry you had to bring up the loss of the Eagles though. Sorry, yes. I

SPEAKER_01: 7:04

apologize.

SPEAKER_02: 7:06

Sami, way in here. What do you think of all this?

SPEAKER_00: 7:10

So I think that coming up with let's just cut the prices so that more people attend the events and we're going to in aggregate get more value is not a sustainable long-term strategy. I think that it's being lazy when everybody is trying to move to personalized pricing and personalized experience It works, I think, in the short term, but when you look at it in a longer horizon over time, I'm not too keen on that because, first of all, differentiated products versus commodity products, I think an event is a differentiated offering. Where you sit at that event, what kind of an experience you have, whether or not your stadium is like the Niner Stadium with super high-tech versus something more kind of rudimentary. These are all things that people are willing to pay or pay more or pay less for. And then the other thing is certain things are perishable goods and their value is really kind of something that changes over time, right? An airplane seat is perishable. A tomato at the supermarket is perishable. A seat at an NFL stadium is perishable. And as the value curve for those events change over time, as you get closer to the event, if there is scarcity around it, the value keeps going up. But the moment that the event starts 15 minutes in, it suddenly has a big drop and it just vanishes. The moment that the airplane door is closed, that empty seat is worth nothing. But 15, 20 minutes earlier, it could have been worth$2,000. So ultimately, price is a value that you capture in the eyes of the consumer. And that's why I think the variable pricing approaches are here to stay. And in economics, we have the whole kind of consumer surplus versus producer surplus thing where something that's worth$100 to me, I will buy at$50 is the consumer surplus. I got a great deal. or something worth, the opposite is the producer surplus where it's in the seller's eyes, it might be worth something like 50 bucks, but people are going to pay a hundred. So you wanna get to the most optimal point where the value matches and the consumer surplus and the producer surplus are maximized to the fullest. So the data science behind variable pricing is huge. People are investing, enormous amounts of money in trying to make it optimal. I actually think that it creates efficiencies because everybody's watching each other. And as long as something is readily available and being produced in ample fashion, that pushes the prices down. If there is a shortage of something, it has the opposite effect, then it starts pushing the prices up. So I guess the upside is for anything that's abundantly available, It reduces the margins for the producers and it makes it most optimal pricing. And the moment that you have the other trend where the demand starts to exceed supply, the prices start to escalate faster than they should. But I believe that it's something that's going to be here to stay for quite some time.

SPEAKER_02: 10:57

Yeah, interesting. As a finance business partner, I want to be obviously delight my customers and maximize profits and cashflow. As a consumer though, you bring up value curve. I wanna pay for what matters most to me, but I also don't wanna be the chump that paid more than everyone else. And I just wonder, is there a risk here that companies could over optimize simply because the data's available, but when consumers talk to each other and they're sort of scratching their heads, why did I have to pay so much more? You

SPEAKER_00: 11:29

know? That's a fantastic point that you're bringing up. That's why what you need to do when you're doing these dynamic pricing models is that you need to treat the group of people who belong to the same cohort with the same principles so that you can maintain a sense of fairness. If you and I are in the same cohort but one of us pay a lot more than the other and there isn't an explainable value viewpoint for that, then that's loses faith in the brand and it can be more expensive than the revenue that the producer captured.

SPEAKER_02: 12:07

Yeah. Interesting and fair point. So Dana, when you're on your NFL stadium tour, you need to keep these things in mind, my friend. I do. Okay. All right. That was interesting. Let's move on to our next topic, which is conversational artificial intelligence. This is commonly used in apps like Siri and Alexa We're starting to see this all over the place. chat GPT has already entered the classroom. It's building restaurant menus. It's writing legal letters. And interestingly enough, possibly infringing on copyright and patents as well. So there's a lot to unpack from a value creation perspective here as well. Sami, you've been called Mr. Data. What's going on in this space and what are the upsides and downsides on value creation that you see?

SPEAKER_00: 13:26

So I think we're at an AI inflection point where these technologies that have been around for some time are becoming mainstream. And as they become mainstream and get in the hands of a lot of different people and a lot of different users, we suddenly start to see the good and the bad and the ugly of all of this. But this didn't happen overnight, right? I mean, this has been in the works. It's been coming up for some time and, you know, essentially large language models. They're pre-trained on massive amounts of text data, and then they're fine-tuned on specific tasks. And this is kind of reminiscent of when we had the internet-enabled applications in the 1990s, where you could just dial up and download some data points, or even things like people handwriting things or hand-typing things to starting to use a copier, which gave you a tremendous jump in productivity. So AI actually has the potential to do that if it gets integrated to different things properly. So fewer people will be able to do more tasks, but also certain things that we started to see as, okay, let's just like put it on a spreadsheet or put it on a document or create a presentation for it. Most of those tasks can start getting automated. So for example, if you deploy a large length large language model within your organization, you can say, hey, based on all the information that we have in our organization, create a report on what's going to happen to such and such in the next like 30 days or make a prediction on this. So it's, you know, really cool stuff, but it does have certain shortcomings. So one, things that you would have assigned a business analyst to do, you know, then they would go disappear for two weeks and come back with some results. That could be automated and get done in like a mere minutes or sometimes maybe a little bit longer than that. On the flip side, these models are super expensive and they don't have accountability. They cannot really necessarily tell you why they came up with the things that they came up with. And it's going to be some time before they become meaningful and they become truly useful. Right now, they're still in the early play stages, playful stages. But I can assure you in the next three, four, five years, it's going to be a ubiquitous part of almost everything that we do.

SPEAKER_02: 16:13

So that's a pretty bullish outlook then. Okay, Dana, what do you say?

SPEAKER_01: 16:18

So I'm sort of in the same boat with Sami. I definitely feel from a business perspective, there's a cost, right? A cost benefit analysis you have to do because you're sort of replacing either a human or tasks with a license and essentially a digital human hire. But that digital human hire may not have the level of experience or the knowledge like If you're running an analysis and you need certain, what I'll call the soft criteria, you know, for extension, you know, I've been in the education business for 10 years. So I know certain things about the industry that might move something higher or lower. A digital asset may actually not have the knowledge for that. So there could inherently be errors in whatever it is that you're trying to run. I think on the consulting side, Sumer side, I would say it's absolutely more helpful. However, there's limitations, right? There's limitations on jargon or slang. I may use certain words in certain phrases, or they may sound different because I'm from New York versus another region. Maybe English is my second language. So if you're dictating verbally to something to do something for you, it might not understand. I think I think, you know, from an Alexa perspective, as a consumer, I asked Alexa a lot of questions that you could probably understand. And Alexa is probably got a hit rate of 50% with me, because she often doesn't understand what I'm asking. And I have to ask the question five different times or shorten the sentence or ask it a different way, or I have to sit down. And by the time I sit down and figure out how I could ask Alexa the question differently, so that she could answer it, I could have just gotten on Alexa. My phone or iPad or laptop, Googled it and gotten the answer I needed. From a business perspective, chats, I'm finding never work. And that just, again, that could be my technology impact. But I find from a business perspective, and a lot of this has to do with banking, never answers my question. I often will use chats to try to get the answer for a question, but my question is so specific to either my company or a specific transaction. It does not answer the question. And then I have to try to find the phone number to call the bank. And then the longer I'm on hold for the bank, the more irritated I get because I just want an answer to my question, which should have been very simple to find. So I have a love-hate relationship with it, but I do believe and I do agree with Sammy that this is the direction that we're going and we just need to get it trained better.

SPEAKER_02: 19:15

Yeah. I'm probably between the two of you on this. It feels like an improvement over the technologies we've already been using. And it's definitely going to create opportunities to further trim costs and improve customer care. I think the natural language models are improving. Are they improving fast enough and working their way into the chatbots and virtual assistants? Not really, because just like you, I can't think of a chatbot experience that has really fully answered my my questions. And we use them heavily actually within my company. Will lawyers stop writing letters and marketeers stop drafting their own ad copy anytime soon? I don't think so. But we will have, and I agree with Sami's timeframe based on what I'm reading and hearing from others, we'll have new tools that will help us frame value creation as we go forward. I don't think it's happened That's my assessment. Does anybody want to weigh in on that?

SPEAKER_00: 20:25

No, I don't think it's happening tomorrow, but think of the spell checker on your document editor. Do you remember the time when you actually had to use a dictionary? It's been a while. Yeah, it's just going to be sprinkled across everything, and it will complete your sentences first, and you'll start writing the first sentence of a paragraph it will finish your paragraph for you. And it will make you super productive, I think, over time.

SPEAKER_02: 21:03

Yeah, and one more question on this. Do you think it will apply beyond roles that tend to be repetitive? Sami, you used the business analyst example. I thought I read something recently. At some point, it could be applied to strategy. So do you need to pay for the big, expensive CEO, for example? Is it routine tasks or could it apply to some of these larger roles as well?

SPEAKER_00: 21:32

You know, I actually think that it's going to excel at routine tasks first. The whole kind of philosophical argument then goes back to whether or not you infer based on correlations or causality, right? A CEO is good because with limited information or sometimes not so limited information, they make a decision with in a timeframe that's the most optimized timeframe. And it involves, if this happens, then this should happen type of a prediction. And it's very much based on, you know, kind of, it's a mixture of gut feeling with data, as opposed to most of these models, at least right now, they don't use gut feeling, they use correlations and very clearly defined rules. So I I think we're some ways away before they can start making the kind of CEO-like strategic decisions. They're good at like obvious decisions.

SPEAKER_02: 22:42

Yeah. I mean, as a finance business partner, I feel like there's two sides to that. And a lot of frontline analysts, for example, make similar decisions, just perhaps at a smaller level. This is interesting. So Dana and Sami, I think we should ask our Why don't we ask our holographic chatbot doppelgangers to handle the next episode? And we're going to charge them each a different host fee, right?

SPEAKER_01: 23:13

Well, it depends upon what day we do it and what time of the day we do it. Exactly. And what the weather outside is. Isn't that right?

SPEAKER_02: 23:18

Absolutely. Absolutely. Yeah. Okay. So I said it was full wonky. It definitely felt like that. It was a lot of fun today. And I want to thank you both for sharing your insights. And to our listeners, there's a ton of value out there just waiting to be created. Keep up the good work and we'll see you next time. Thank you, Matt. Thank you, Dana.

SPEAKER_01: 23:39

Thank you both.